Original subtitles

OK, it really wants to pitch down.

The pilots of an MD-83 suddenly lose control.

The pilots of an MD-83 suddenly lose control.

They need to find a way to control

the pitch of the jet.

the pitch of the jet.

Without warning, an ATR-72 rolls to the right.

Without warning, an ATR-72 rolls to the right.

That airplane can fall out from under them

in the blink of an eye.

in the blink of an eye.

Two Boeing 737s suffer the same deadly fate

after going into sudden dives.

after going into sudden dives.

If a pilot cannot control the pitch system,

If a pilot cannot control the pitch system,

literally, that means the pilot cannot control the plane.

These are the toughest cases to solve...

These are the toughest cases to solve...

speeding planes diving nose first into the ground.

There was very little left of the airplane.

There was very little left of the airplane.

The plane pitched full nose trim down.

Investigators in three cases hunt

Investigators in three cases hunt

for answers when pilots lose control of their plane's pitch.

It just jammed, stopped working completely.

It just jammed, stopped working completely.

Take a look.

I never allowed myself to think this investigation

I never allowed myself to think this investigation

could go undetermined.

Ladies and gentlemen, we are starting our approach.

We lost both engines.

We lost both engines.

Put the mask over your nose.

Emergency descent.

Mayday, Mayday.

Brace for impact!

It's gonna crash!

Alaska Airlines Flight 261 climbs into sunny skies

over Puerto Vallarta, Mexico, carrying

over Puerto Vallarta, Mexico, carrying

83 passengers and five crew.

Gear up.

Gear up.

Gear is up.

Captain Ted Thompson is an Air Force veteran.

Captain Ted Thompson is an Air Force veteran.

Thank you, sir.

First officer Bill Tansky has been

First officer Bill Tansky has been

flying for almost four decades.

The pilots on 261 were highly regarded.

The pilots on 261 were highly regarded.

These were really top-notch aviators.

The MD-83 will be stopping over in San Francisco

The MD-83 will be stopping over in San Francisco

before heading to its final destination,

Seattle, Washington.

Seattle, Washington.

But 15 minutes after taking off, there's a problem.

But 15 minutes after taking off, there's a problem.

It's pushing down.

The nose of the plane is dropping.

The nose of the plane is dropping.

And First Officer Tansky must wrestle with the control column

to keep the plane climbing.

to keep the plane climbing.

Increasingly, there's more and more pressure

that he's having to hold.

All right.

Well, we're still climbing.

Well, we're still climbing.

21 minutes into the flight,

they level off at 31,000 feet.

they level off at 31,000 feet.

But it appears the stabilizer is jammed.

No dice.

Let's try it on autopilot.

Let's try it on autopilot.

Plane is steady.

The autopilot is keeping the plane

The autopilot is keeping the plane

at the correct altitude, but Captain Thompson has no idea

how long that will continue.

how long that will continue.

All right.

I'm thinking we get this plane on the ground

sooner rather than later.

sooner rather than later.

He's worried about the stabilizer,

so he decides to divert to Los Angeles International Airport.

so he decides to divert to Los Angeles International Airport.

Dispatch 261, requesting a diversion to LAX.

Our longitudinal trim system is inoperative.

Our longitudinal trim system is inoperative.

Thompson contacts Alaska

Airlines dispatch in Seattle.

Airlines dispatch in Seattle.

Dispatch 261, copy that.

This will click the autopilot off.

This will click the autopilot off.

Stabilizer motion.

Stabilizer motion.

Stabilizer motion.

Holy crap.

Holy crap.

Flight 261 goes into a dive.

It got worse.

Center, Alaska 261.

We are in a dive here.

I've lost control of vertical pitch.

I've lost control of vertical pitch.

Alaska 261, say again, sir?

We are in a vertical dive at 26,000.

We are in a vertical dive at 26,000.

Speed brakes.

The speed brakes are movable surfaces on the wings

The speed brakes are movable surfaces on the wings

that increase drag.

The maneuver works.

Once we get the speed slowed, maybe we'll be OK.

Once we get the speed slowed, maybe we'll be OK.

The plane has dropped 8,000 feet in just 80 seconds.

The plane has dropped 8,000 feet in just 80 seconds.

We got it back.

Under control here.

Under control here.

No, we don't.

It still requires all their strength to keep

It still requires all their strength to keep

the nose of the plane level.

OK.

OK.

It really wants to pitch down.

They're in a fight for control of the airplane.

The passengers' anxiety grows.

The passengers' anxiety grows.

We better talk to the people back there.

Yeah, I know.

Yeah, I know.

Folks, we've had a flight control problem.

We're intending to go to Los Angeles.

We're intending to go to Los Angeles.

We're working on it, and I don't anticipate

any problems once we get a couple systems back on the line.

any problems once we get a couple systems back on the line.

As they near Los Angeles,

the pilots prepare for an emergency landing.

the pilots prepare for an emergency landing.

OK, we're pretty stable here.

But we got to get down to 180 knots.

But we got to get down to 180 knots.

They need to find a way to control

the pitch of the jet.

the pitch of the jet.

If it's controllable, we ought to just try and land it.

They would not be able to pull

They would not be able to pull

the plane up and do a go-round.

They had one shot to get it on the ground or all would be lost.

They had one shot to get it on the ground or all would be lost.

You feel that?

You feel that?

Yeah.

But at that moment, disaster strikes.

But at that moment, disaster strikes.

The airplane pitched down and rolled.

This is a pilot's nightmare.

Mayday!

Mayday!

Push and roll.

Push and roll.

Push and roll.

Flight 261 continues to roll left and dive.

Flight 261 continues to roll left and dive.

Pull up.

Pull brake.

Pull brake.

And this is like an air show maneuver.

OK, we are inverted.

OK, we are inverted.

Now we got to get it up.

They've got to get the airplane back right side up.

They've got to get the airplane back right side up.

Their one hope is that they

could kick that rudder, they could flip the plane back over.

could kick that rudder, they could flip the plane back over.

OK, let's kick rudder.

Left rudder, left rudder.

Left rudder, left rudder.

The pilots are kind of hanging

upside down like bats.

And it was very difficult to reach the rudder pedal.

And it was very difficult to reach the rudder pedal.

I can't reach it.

OK.

We gotta get it over again.

We gotta get it over again.

Speed brakes.

Speed brakes.

They were giving everything they had.

The windscreen's full of the ocean.

They're not gonna solve this.

They're not gonna solve this.

Ah, here we go!

Rescuers search for survivors.

Rescuers search for survivors.

They find none.

All 88 passengers and crew on board Flight 261 are dead.

Investigators from the National Transportation Safety Board,

Investigators from the National Transportation Safety Board,

the NTSB, are assigned to find an explanation for the crash.

the NTSB, are assigned to find an explanation for the crash.

LAX is saying the pilots reported a jammed stabilizer.

We knew quite a lot just

from the transmissions between the flight crew and Los Angeles.

from the transmissions between the flight crew and Los Angeles.

The plane pitched full nose trim down.

The plane pitched full nose trim down.

We immediately began

to research the horizontal stabilizer trim system.

to research the horizontal stabilizer trim system.

The leading edge of the horizontal stabilizer

The leading edge of the horizontal stabilizer

is raised or lowered by a jackscrew.

The jackscrew changes the stabilizers angle as it

The jackscrew changes the stabilizers angle as it

moves through an acme nut.

We need to see the jackscrew assembly.

We need to see the jackscrew assembly.

Investigators wonder how the horizontal trim system

could have failed.

could have failed.

I'll get this to the Navy.

Recovering parts of the stabilizer system

could give investigators important clues.

could give investigators important clues.

But the wreckage field lies in the bottom of the ocean,

But the wreckage field lies in the bottom of the ocean,

beyond the reach of divers.

Nine days into the investigation, remotely operated

vehicles equipped with robotic arms

vehicles equipped with robotic arms

retrieve the parts investigators are looking for.

We got the jackscrew.

We got the jackscrew.

With the brainpower we had, we had

a fighting chance to find out exactly what

went wrong during this flight.

went wrong during this flight.

Systems investigator

Jeff Guzzetti immediately spots a problem

Jeff Guzzetti immediately spots a problem

with the jackscrew assembly.

Why isn't the nut attached to the jackscrew?

When we looked at that,

we asked ourselves, how can that happen?

we asked ourselves, how can that happen?

Threads inside the nut should hold the jackscrew

securely in place.

securely in place.

But these threads failed.

Incredible.

Incredible.

It's completely stripped.

No one thought that you could get

No one thought that you could get

those thick threads to rip out.

Metallurgist Joe Epperson

examines the stripped nut and the jackscrew.

examines the stripped nut and the jackscrew.

There's some grease here on the bottom.

The way to prevent wear

The way to prevent wear

is by lubrication, with grease.

In flight, the jackscrew

In flight, the jackscrew

rotates inside the nut.

To prevent wear, it needs to be lubricated frequently.

To prevent wear, it needs to be lubricated frequently.

In the case of an extreme wear event,

you want to look at is there grease

you want to look at is there grease

where it's supposed to be?

There should be more grease here in the middle.

There should be more grease here in the middle.

There was very minimal signs that there

was any grease on it at all.

was any grease on it at all.

During maintenance,

grease is applied to the interior of the nut

grease is applied to the interior of the nut

through a small valve called a Zerk fitting,

but the valve is blocked.

but the valve is blocked.

So the Zerk fitting is this fitting right here.

So the Zerk fitting is this fitting right here.

And what mechanics have to do is put a grease gun hose in here

and then squirt grease into this little grease fitting

and then squirt grease into this little grease fitting

that goes inside this passage.

So that Zerk fitting Should have preserved and kept the remnants

So that Zerk fitting Should have preserved and kept the remnants

of whatever grease it saw last.

It's packed with dried grease.

It's packed with dried grease.

How long has it been clogged like this?

At least a year, maybe more.

I've been around grease long enough to know

I've been around grease long enough to know

that if you don't replenish it and if you just leave it,

eventually it dries up and gets hard.

eventually it dries up and gets hard.

And it basically says that it had not been replenished.

And it basically says that it had not been replenished.

I'd say we have a maintenance issue here.

NTSB investigators tracked down the mechanic who

NTSB investigators tracked down the mechanic who

serviced the jackscrew and acme nut on the Alaska Airlines MD-83

four months before it crashed.

four months before it crashed.

Could you walk us through how you lubricate

the jackscrew assembly?

the jackscrew assembly?

We were very suspicious about how well

this component was lubricated.

this component was lubricated.

We use a grease gun to the Zerk fitting.

Well, how do you know

whether the lubrication is being done properly

whether the lubrication is being done properly

and when to stop pumping the grease gun?

I don't.

I don't.

He couldn't have known that that Zerk

fitting has taken grease.

You have no idea that it's clogged.

But it's not just the way

the jackscrew assembly was lubricated

the jackscrew assembly was lubricated

that bothers investigators.

The last time the jackscrew was lubricated

The last time the jackscrew was lubricated

was about four months before the accident.

And before that?

And before that?

January 1999.

They're doing it every 2,500 hours.

They're doing it every 2,500 hours.

They discover longer and longer

They discover longer and longer

periods between lubrications.

Check it out.

Check it out.

The airline made multiple requests to extend the intervals

on the lubrications.

In 1987, the interval between lubrications was 500 hours.

In 1987, the interval between lubrications was 500 hours.

Those intervals are measured by the number of hours

Those intervals are measured by the number of hours

the plane is in the air.

In 1991, it goes up to 1,200 hours.

In 1991, it goes up to 1,200 hours.

In 1991, it goes up to 1,200 hours.

By 1996, it increases all the way up to 2,500 hours.

By 1996, it increases all the way up to 2,500 hours.

If you're gonna extend these lubrication functions,

If you're gonna extend these lubrication functions,

then you better be doing something to make sure that what

you're doing is correct.

you're doing is correct.

But even if the lubrication wasn't being

done properly, regular inspections should have caught

done properly, regular inspections should have caught

the wear on that acme nut.

Was the jackscrew assembly on Flight 261

Was the jackscrew assembly on Flight 261

inspected when and how it should have been?

We looked at the maintenance

We looked at the maintenance

records for information about the last check.

Yeah, this doesn't look right to me.

Yeah, this doesn't look right to me.

The team finds paperwork showing

that a mechanic determined the acme nut was

that a mechanic determined the acme nut was

badly worn more than two years before the crash.

Are you sure the reading is 0.040?

Are you sure the reading is 0.040?

The lead mechanic ordered the nut to be replaced.

This was evidence that someone had

This was evidence that someone had

caught the fact that this acme nut was worn out.

Most airlines, the decision would be,

Most airlines, the decision would be,

we will get the piece to the airplane as soon

as possible and replace it.

as possible and replace it.

Do you remember inspecting an MD-83 on September 27, 1997?

I remember it well.

I remember it well.

Investigators contact John Liotine,

the mechanic who reported the worn acme nut

the mechanic who reported the worn acme nut

on Flight 261.

I wrote up the evaluation.

I wrote up the evaluation.

"The nut is worn down.

Replace it."

But when investigators probe further,

they learn that another mechanic canceled

they learn that another mechanic canceled

the order to replace the nut.

That mechanic determined that it was still within safe limits

That mechanic determined that it was still within safe limits

and cleared the plane to fly.

How could a maintenance facility allow

How could a maintenance facility allow

the airplane to be put back into revenue service with the wear

the airplane to be put back into revenue service with the wear

that it found on the acme nut?

This accident could have been prevented.

Literally $0.50 worth of grease cost a planeload

Literally $0.50 worth of grease cost a planeload

of people their lives.

In the wake of the investigation,

federal regulators reduced the interval between jackscrew

federal regulators reduced the interval between jackscrew

lubrications at Alaska Airlines from 2,500 hours to 650 hours.

lubrications at Alaska Airlines from 2,500 hours to 650 hours.

Looking back, it makes no sense at all.

They decided you could put off maintenance, put off maintenance

They decided you could put off maintenance, put off maintenance

and put off maintenance, with the approval, I might add,

and put off maintenance, with the approval, I might add,

of the authorities and the airline who

allowed greasing their parts to be extended almost indefinitely.

allowed greasing their parts to be extended almost indefinitely.

Poor maintenance can play havoc with even

Poor maintenance can play havoc with even

the best designed aircraft.

But sometimes, flaws can develop while a plane

But sometimes, flaws can develop while a plane

is still on the drawing board.

There are 64 passengers on board American Eagle Flight 4184

There are 64 passengers on board American Eagle Flight 4184

as it approaches Chicago's O'Hare Airport.

as it approaches Chicago's O'Hare Airport.

Chicago center, we have discretion down to one

Chicago center, we have discretion down to one

zero thousand.

We're on our way down now.

Eagle 184, Roger.

Fog and reduced visibility at O'Hare

Fog and reduced visibility at O'Hare

are jamming up traffic and creating delays.

are jamming up traffic and creating delays.

Eagle 184, hold southeast on Victor seven.

Chicago center, roger.

Chicago center, roger.

Hold southeast on Victor seven.

Eagle 184.

Captain Orlando Aguilar is in command of the ATR-72.

Captain Orlando Aguilar is in command of the ATR-72.

He programs the autopilot to keep

the plane circling at 10,000.

the plane circling at 10,000.

I just think it's a real high deck

angle in these turns.

Yeah.

The circling turn keeps

The circling turn keeps

the plane's nose pitched up.

You want flaps 15.

That'll bring the nose down.

That'll bring the nose down.

First Officer Jeff Gagliano

extends the flaps to 15 degrees, and the aircraft levels out.

extends the flaps to 15 degrees, and the aircraft levels out.

That's much nicer now, flaps 15.

Sure.

Sure.

Once they let us out of the hold and forget they're down,

we'll get the overspeed warning.

A half hour later, air traffic control directs

the flight to a lower altitude.

the flight to a lower altitude.

Eagle Flight 184, descend and maintain 8,000.

Down to 8,000, Eagle 184.

Down to 8,000, Eagle 184.

Eagle Flight 184, should be about 10

Eagle Flight 184, should be about 10

minutes till you're cleared in.

Thank you.

Thank you.

Are we out of the hold?

No, just going down to 8,000.

No, just going down to 8,000.

I knew we'd do that.

The overspeed warning sounds.

The overspeed warning sounds.

I was trying to keep it at 180.

The plane is going too fast to fly with extended flaps.

The plane is going too fast to fly with extended flaps.

Gagliano retracts them.

Suddenly, the plane banks sharply to the right.

Suddenly, the plane banks sharply to the right.

The control column is stuck.

American Eagle Flight 4184 is in an uncontrolled dive.

American Eagle Flight 4184 is in an uncontrolled dive.

The pilots fight to pull out of it and level off.

The pilots fight to pull out of it and level off.

See? That's it.

Pull up.

Pull up.

Nice and easy.

Pull up.

Terrain.

Pull up.

Flight 4184 crashes in a field

Flight 4184 crashes in a field

just outside Roselawn, Indiana.

All 68 people on board are killed.

All 68 people on board are killed.

The next morning, an NTSB team is on the scene.

It was one of those types of accidents where there was very

It was one of those types of accidents where there was very

little left of the airplane.

Amidst the destruction,

the airplane's two flight recorders are found intact.

the airplane's two flight recorders are found intact.

While the voice and data recordings are analyzed,

investigators focus on the holding

investigators focus on the holding

pattern the flight was put into by air traffic control.

Eagle 184, hold southeast on Victor seven.

Well, folks, I do regret to inform you

Well, folks, I do regret to inform you

that air traffic control has put us into a holding pattern

up here.

When we looked at the air traffic control records,

When we looked at the air traffic control records,

we found that the flow rate, that is the number of airplanes

that were going into Chicago O'Hare at the time,

that were going into Chicago O'Hare at the time,

had to be reduced because of the weather moving through the area.

Chief investigator Greg Feith

Chief investigator Greg Feith

wants to know more about the weather during the hold.

I want weather data, pilot reports,

atmospheric conditions, Doppler radar, anything you can find.

atmospheric conditions, Doppler radar, anything you can find.

Pilots of other planes circling

O'Hare Airport that day point investigators

O'Hare Airport that day point investigators

toward an important clue.

They all experienced icing, some as much as 3/4 of an inch.

They all experienced icing, some as much as 3/4 of an inch.

Icing is a well-known threat to all aircraft.

Icing is a well-known threat to all aircraft.

It creates drag.

It reduces lift.

It causes performance and handling problems.

It causes performance and handling problems.

Examining weather records from the day

of the crash, investigators discover that during

of the crash, investigators discover that during

the 39-minute holding pattern, Flight 4184 repeatedly

the 39-minute holding pattern, Flight 4184 repeatedly

passed through cloud banks.

4184 we knew had been holding in an icing event.

There were layers of clouds that they

There were layers of clouds that they

had been flying in and out of.

But did the pilots know they were

But did the pilots know they were

flying into icy conditions?

The cockpit voice recorder provides the answer.

The cockpit voice recorder provides the answer.

Oh, that's much nicer now, flaps 15.

Yeah.

Showing some ice now.

I'm sure now we'll get the overspeed warning.

Nine minutes before the crash,

Nine minutes before the crash,

they noticed their plane was picking up ice.

Feith and his team wonder what the crew did about it.

Feith and his team wonder what the crew did about it.

They turn to the flight data recorder.

They turn to the flight data recorder.

It indicates that almost 17 minutes before the crash,

the master caution warning went off.

the master caution warning went off.

In response, Captain Aguilar turned the plane's de-icing

system to its maximum setting.

system to its maximum setting.

But was the de-icing system working?

But was the de-icing system working?

Normally, a rubber boot inflates to break

ice that coats the surface of the leading edge of the wing.

ice that coats the surface of the leading edge of the wing.

The key component to the de-icing system

The key component to the de-icing system

is probably not the black rubber boot that you see

on the airplane, but the timer and the valves

on the airplane, but the timer and the valves

that control those boots.

The force of the crash has

left markings that show the position of the valves

left markings that show the position of the valves

at the time of impact.

Those markings indicate the plane's de-icing system

Those markings indicate the plane's de-icing system

was working properly.

Showing some ice now.

So if there wasn't a problem

with the de-icing system, what else

with the de-icing system, what else

could have caused the crash?

Other pilots believe they know the answer and decide to act.

Other pilots believe they know the answer and decide to act.

I distributed a brochure

which essentially warned people about the problems

with the ATR aircraft.

with the ATR aircraft.

The brochure says the ATR-72 fleet operated

by American Eagle is dangerous in icing conditions

by American Eagle is dangerous in icing conditions

and the airline is doing nothing about it.

and the airline is doing nothing about it.

A month after the crash, pilot frustration boils over.

A group of a dozen pilots refuses to fly the ATR,

A group of a dozen pilots refuses to fly the ATR,

citing risky weather.

The public protest leads investigators

The public protest leads investigators

to take a closer look at the aircraft's vulnerability to ice.

to take a closer look at the aircraft's vulnerability to ice.

The NTSB Charlie Perera visits the ATR's French manufacturer,

Aerospatiale.

Aerospatiale.

We embarked on a research effort

to try to identify all previous ATR-42

to try to identify all previous ATR-42

and 72 roll control incidents.

and 72 roll control incidents.

He finds a disturbing pattern.

And about five...

And about five...

five of those were found to be similar after the fact in our

review to the Roselawn case.

review to the Roselawn case.

In December 1988, an American Eagle ATR-42,

In December 1988, an American Eagle ATR-42,

preparing to land in icy conditions,

preparing to land in icy conditions,

suddenly pitched down and stalled without warning.

Terrain, terrain.

Terrain, terrain.

Terrain, terrain.

The pilots recovered by boosting engine power.

But other flights end differently.

But other flights end differently.

In October 1987, an ATR-42 en route to Germany

In October 1987, an ATR-42 en route to Germany

crashed, killing all 37 people on board.

crashed, killing all 37 people on board.

The investigation found that ice was a factor,

but placed the blame on pilot procedures.

but placed the blame on pilot procedures.

For investigators, it's really important

to look back at historical information about any airplane.

to look back at historical information about any airplane.

Did any of your studies uncover a wing design flaw?

Charley Pereira wants to know if the engineers who

Charley Pereira wants to know if the engineers who

built the ATR series can explain why the plane

built the ATR series can explain why the plane

would go out of control.

What he's told...

Still got ice.

Still got ice.

...will shock the airline industry.

NTSB investigator Charley Pereira

presses engineers for the reasons

presses engineers for the reasons

behind the ATR's tragic record.

You've seen the flight data from the accident flight.

You've seen the flight data from the accident flight.

Where would the ice have formed to cause that?

When I asked one of their oldest, most senior

When I asked one of their oldest, most senior

aerodynamicist what he thought would cause the behavior that we

saw on the flight recorder for the Roselawn airplane,

saw on the flight recorder for the Roselawn airplane,

he drew a little shape just behind the de-ice boots.

he drew a little shape just behind the de-ice boots.

Aft of the de-icing boot.

If ice builds up behind the boot,

the crew has no way to clear it.

We still got ice.

But how could ice form in that location?

Additional weather analysis shows

Flight 4184 flew into something much more dangerous

Flight 4184 flew into something much more dangerous

than freezing rain.

Supercooled large droplets, or SLDs,

raindrops the size of a human hair that stay in liquid form

raindrops the size of a human hair that stay in liquid form

even when the temperature is below freezing.

They only freeze when they contact a solid.

On every lap of its holding pattern,

On every lap of its holding pattern,

Flight 4184 flew through layers of cloud that contained

these supercooled droplets.

these supercooled droplets.

A warm cloud top minus 8 to minus 12 degrees C,

we had drizzle formation near the top

we had drizzle formation near the top

that was probably encouraged by some wind

shear near the cloud top.

They just put that poor guy in the wrong place

They just put that poor guy in the wrong place

at the wrong time.

Investigators learn that while ordinary ice

Investigators learn that while ordinary ice

accumulates on the leading edge of the wing,

where de-icing boots can get rid of it,

where de-icing boots can get rid of it,

SLDs slide back over the boots before freezing

and form a ridge out of reach of a de-icing system

and form a ridge out of reach of a de-icing system

and beyond the view of the flight crew.

Eagle Flight 184, descend and maintain 8,000.

Eagle Flight 184, descend and maintain 8,000.

Down to 8,000., Eagle 184.

Down to 8,000., Eagle 184.

Investigators conclude that the ice buildup had no

effect on Flight 4184 until the plane

effect on Flight 4184 until the plane

began its descent to 8,000.

I knew we'd do that.

I knew we'd do that.

The descent increased the plane's airspeed

and triggered a warning from the flight computer,

telling the pilots they were flying too

telling the pilots they were flying too

fast with the flaps extended.

I was trying to keep it at 180.

I was trying to keep it at 180.

The crew then retracted the flaps,

but that routine action had a dramatic effect

but that routine action had a dramatic effect

on the plane's aerodynamics.

When they retracted the flaps as they started

When they retracted the flaps as they started

their descent down to 8,000 feet,

the airplane pitched back up into a very nose-high attitude.

When the nose lifted, air flow over the ice-laden wing

When the nose lifted, air flow over the ice-laden wing

was disrupted.

This created powerful turbulence,

lifting the aileron in the right wing

lifting the aileron in the right wing

and causing the plane to roll.

The forces were determined

to be about 250 pounds.

to be about 250 pounds.

So for the pilot, it would have been trying to roll against 250

pounds of aerodynamic force.

pounds of aerodynamic force.

Damn!

The NTSB has uncovered a deadly flaw

in a popular airplane.

The ATR's wing design makes it vulnerable to icing.

The ATR's wing design makes it vulnerable to icing.

In response, the plane's French builder, Aerospatiale,

makes the de-icing boots wider, so they cover

makes the de-icing boots wider, so they cover

a larger part of the wing.

Several things had to go wrong for this to happen.

Several things had to go wrong for this to happen.

And it was only because of the very detailed investigation

in these accidents that these were able to be resolved so

in these accidents that these were able to be resolved so

that it would not happen again.

It took more than a year

to wrap up the investigation into Flight 4184,

to wrap up the investigation into Flight 4184,

but the factors involved in other planes' pitch problems...

but the factors involved in other planes' pitch problems...

- Pull! - Oh, god.

Pull!

...can be so rare and unusual that it

takes several years to solve the mystery.

takes several years to solve the mystery.

United Airlines Flight 585 is on final approach

into Colorado Springs with 20 passengers

into Colorado Springs with 20 passengers

and five crew on board.

The skies are clear, but heavy turbulence

The skies are clear, but heavy turbulence

is rocking the flight.

Nice-looking day.

Hard to believe the skies are unfriendly.

Hard to believe the skies are unfriendly.

At the controls is 52-year-old

Captain Harold Green.

Green's co-pilot is Patricia Eidson.

Flight attendants, prepare for landing.

Flight attendants, prepare for landing.

At Colorado Springs Municipal Airport,

controllers are ready to bring Flight 585 in.

controllers are ready to bring Flight 585 in.

United 585 is cleared for a visual approach to runway 35.

Starting on down.

Starting on down.

Suddenly, the 737 spins out of control.

Suddenly, the 737 spins out of control.

What's going on?

Oh my god!

Oh my god!

Oh my god!

Flight 585 is in a 230 mile an hour dive.

Flight 585 is in a 230 mile an hour dive.

It crashes less than four miles from the runway.

The shattered remains of United Airlines Flight 585

lie buried in a fire-blackened impact crater.

lie buried in a fire-blackened impact crater.

Rescue workers arrive within minutes.

Rescue workers arrive within minutes.

None of the 25 people on board have survived.

None of the 25 people on board have survived.

By nightfall, NTSB investigators are on site.

My first sense that it was going to take some time

to investigate the accident was the damage

to investigate the accident was the damage

that we saw in the parts.

When they're burnt and broken, the process always takes longer.

When they're burnt and broken, the process always takes longer.

The black boxes are extracted from the site

and sent to the lab in Washington.

and sent to the lab in Washington.

While poring over the voice recordings,

human performance specialist Malcolm Brenner

human performance specialist Malcolm Brenner

hears nothing to indicate that the pilots

struggled with the approach.

struggled with the approach.

This has the sense of an excellent crew caught randomly,

if anything.

if anything.

So again, that was my first impression,

is that this would be consistent more with a hardware situation.

is that this would be consistent more with a hardware situation.

Investigators examine the plane's

wings and tail assembly.

wings and tail assembly.

After eliminating other flight control surfaces that we thought

After eliminating other flight control surfaces that we thought

could contribute to the roll, we started looking at the rudder.

The violence of the crash

The violence of the crash

makes recovering these parts a daunting challenge.

Almost nothing left.

But a vital component of the tail assembly

But a vital component of the tail assembly

is mostly intact, the power control unit, or PCU.

When a pilot pushes on a rudder pedal,

the PCU uses hydraulic fluid to convert

the PCU uses hydraulic fluid to convert

movements into pressure needed to move the 737's rudder.

movements into pressure needed to move the 737's rudder.

Inside the PCU, a dual servo valve

directs the flow of pressurized hydraulic fluid

directs the flow of pressurized hydraulic fluid

that moves the rudder.

When a technician opens the power control unit,

When a technician opens the power control unit,

chips of metal are found floating in the hydraulic fluid.

It's an eye-opening find.

It's an eye-opening find.

Could these particles have caused the servo valve to jam?

Could these particles have caused the servo valve to jam?

Testing produces no evidence that they

led to the loss of control.

We didn't have any absolute indication or information

We didn't have any absolute indication or information

that we could point to that said the rudder power control

unit, the servo valve, or any part of that flight control

unit, the servo valve, or any part of that flight control

system caused that accident.

Over the next 21 months,

Over the next 21 months,

investigators rule out pilot error, mechanical failure,

and weather.

and weather.

Then, for only the fourth time in its history,

the NTSB releases a report which doesn't identify a cause.

the NTSB releases a report which doesn't identify a cause.

We had put a lot of time and effort into the investigation,

We had put a lot of time and effort into the investigation,

and we just weren't sure what had happened.

Two years later, US Air Flight 427

approaches Pittsburgh.

It hits a patch of turbulence.

It hits a patch of turbulence.

Whoa, Jesus.

The 737 suddenly rolls left.

The 737 suddenly rolls left.

Hold on, hold on.

Hold on, hold on.

Hold on.

Nothing the pilots do can

pull the plane out of the dive.

pull the plane out of the dive.

What the hell is this?

Shoot.

Shoot.

427, emergency!

427, emergency!

Oh, shoot.

Pull.

Oh, god.

Pull!

Pull!

Oh, god.

No!

First responders arrive quickly,

First responders arrive quickly,

but there's no hope for the 132 passengers and crew.

NTSB investigators arrive at the site

of a second Boeing 737 crash.

of a second Boeing 737 crash.

There was no aircraft there.

There were only bits and pieces of the airplane.

There were only bits and pieces of the airplane.

It wasn't really recognizable as an airplane.

Eyewitness accounts, along with voice and data

Eyewitness accounts, along with voice and data

from the flight recorders, indicate Flight 427

suddenly rolled and then dived.

suddenly rolled and then dived.

It matches what happened aboard United 585.

It matches what happened aboard United 585.

Certainly the whole team was aware of the previous accident

Certainly the whole team was aware of the previous accident

with United 585 in Colorado Springs.

We tried to keep that in the back of our minds

and take a look at this one as to what it presents to us.

and take a look at this one as to what it presents to us.

Almost immediately, investigators

make a promising discovery.

make a promising discovery.

Much of US Air 447's tail and rudder appear intact.

The hydraulic devices inside the tail

have also sustained very little damage.

have also sustained very little damage.

Careful with it.

Careful with it.

The primary suspect is, again, the dual servo valve.

Just like in the earlier crash, tiny chips of metal

Just like in the earlier crash, tiny chips of metal

are found floating in the hydraulic fluid.

But once again, there's no evidence the debris interfered

But once again, there's no evidence the debris interfered

with the delicate valve.

There wasn't any indication that it had failed.

There wasn't any indication that it had failed.

And it operated within the parameters we expected it to.

We were all frustrated as months wore into years.

We were all frustrated as months wore into years.

What were we missing?

What were we missing?

Investigators need a break in the case.

On June the 9th, 1996, they get one.

On June the 9th, 1996, they get one.

An Eastwind 737 is on final approach

An Eastwind 737 is on final approach

to Richmond, Virginia, when, without warning, it

rolls sharply to the right.

We knew we had a problem with the rudder.

We knew we had a problem with the rudder.

I turned the yoke the opposite direction,

stood on the opposite rudder pedal.

The pedal didn't move for me.

The pedal didn't move for me.

Then suddenly, the unknown forces holding the jet

let go.

let go.

In a matter of seconds, it released

itself and went back to normal.

itself and went back to normal.

The Eastwind 737 lands safely.

The next day, the NTSB team arrives in Richmond.

The airplane literally didn't move.

It stayed in its location in the airport until we got down there.

It stayed in its location in the airport until we got down there.

It gave the NTSB a tremendous break

because suddenly they had a 737 that had had a rudder

because suddenly they had a 737 that had had a rudder

incident that was intact.

And they had a pilot who was alive

And they had a pilot who was alive

and who could talk about it.

NTSB investigators quickly

determined that what happened on board

determined that what happened on board

Eastwind Flight 517 is alarmingly similar to events

on Flights 427 and 585.

on Flights 427 and 585.

They questioned the Eastwind pilots.

They questioned the Eastwind pilots.

And when we said, well, what happened?

they said there was something wrong with the rudder pedal.

The pedal wouldn't go down.

The pedal wouldn't go down.

I was standing on the rudder pedal,

and I couldn't get it to go down.

My god.

Investigators immediately

focused their attention on Eastwind's rudder controls.

focused their attention on Eastwind's rudder controls.

The PCU is removed, inspected, and tested again and again.

To everyone's frustration, the unit performs perfectly.

We tested that aircraft as-is.

It was intact.

It was intact.

We went through it completely.

We did flight tests with it.

And it passed all tests.

And it passed all tests.

Investigators go back to the control unit

from US Air Flight 427 and put it through more extreme testing.

from US Air Flight 427 and put it through more extreme testing.

One fellow mentioned a test they had done in the military

One fellow mentioned a test they had done in the military

of a thermal shock, where if you had the actuator being very cold

of a thermal shock, where if you had the actuator being very cold

and put in very hot hydraulic fluid,

it would cause it to react in strange ways.

it would cause it to react in strange ways.

So we put together a thermal shock test.

So we put together a thermal shock test.

The PCU is super cooled

to simulate the minus 40 degree temperatures at 30,000 feet.

to simulate the minus 40 degree temperatures at 30,000 feet.

Then it's quickly injected with superheated hydraulic fluid.

As we were standing there listening to the actuator

As we were standing there listening to the actuator

move left and right, left and right, it stopped, and it

move left and right, left and right, it stopped, and it

was not commanded to stop.

It just jammed, stopped working completely.

It just jammed, stopped working completely.

It appears that the small hydraulic valve that

It appears that the small hydraulic valve that

controls the rudder of the Boeing 737

can jam in certain circumstances.

can jam in certain circumstances.

But that's not all.

The most important breakthrough

came when a Boeing engineer examining the data

came when a Boeing engineer examining the data

from that test discovered some numbers

from that test discovered some numbers

that indicated the valve, at that point,

had actually reversed.

had actually reversed.

It is a stunning revelation.

Hold on.

Hold on.

Not only can the dual servo valve jam,

but it can also cause the rudder to reverse.

You turn to the right, it goes left.

That's what the pilots were faced with,

something so unusual that they didn't

understand what was happening.

To the very end, the pilots of Flight 427

had pushed hard, hoping the rudder would help pull

had pushed hard, hoping the rudder would help pull

them out of a deadly spiral.

But they had no way of knowing they were making the plane's

But they had no way of knowing they were making the plane's

drastic roll even worse.

No!

Flight attendants, prepare for landing.

Flight attendants, prepare for landing.

Reevaluating the data of United Flight 585,

Reevaluating the data of United Flight 585,

investigators conclude its flight crew fell

victim to the same malfunction.

victim to the same malfunction.

Going back to Colorado Springs,

you could follow a progression of what the captain was doing.

He's close to the ground and suddenly, under rudder reversal,

he puts in a little bit of pedal.

The pedal violently pushes his leg back.

The pedal violently pushes his leg back.

All down.

15 flap.

15.

Rudder reversal certainly fits what I know about this crew.

We were able to show the failure mode.

We were able to show the failure mode.

It matched the flight data recorder from each aircraft.

It fit like a glove.

It fit like a glove.

Hold on.

Oh, my god!

It took less than 10 seconds to go

from rudder reversal to impact.

from rudder reversal to impact.

In the aftermath of the investigation,

sweeping changes are made to improve

sweeping changes are made to improve

the safety of the Boeing 737.

The federal regulator directs Boeing to redesign

The federal regulator directs Boeing to redesign

the rudder's dual servo valve to eliminate the potential

for reversal and to replace those valves on thousands

for reversal and to replace those valves on thousands

of 737s around the world.

of 737s around the world.

One thing we don't like at the Safety Board

is to have an undetermined accident,

because then we can't make a change to improve safety.

because then we can't make a change to improve safety.

So out of US Air 427, United 585,

we have a much safer 737 fleet.

we have a much safer 737 fleet.

It took the NTSB 10 years

It took the NTSB 10 years

to solve the mysterious crashes of Flights

585 and 427, the longest investigation in its history.

585 and 427, the longest investigation in its history.

You know, the first thing people say is, oh,

You know, the first thing people say is, oh,

who would have thought this could happen?

But when they go back and look at the data, often it's there.

It was a problem waiting to happen.

It was a problem waiting to happen.

A pitch problem can cause catastrophe so total

A pitch problem can cause catastrophe so total

that evidence is scarce.

But by revealing deeply hidden flaws,

these three grueling investigations changed aviation

these three grueling investigations changed aviation

and made the skies safer.

and made the skies safer.

The biggest lesson to take away for the industry

and for the public at large is that in a complex accident,

and for the public at large is that in a complex accident,

such as this, the answers may not come in a few months or even

a few years.

Sometimes, it will take several events

Sometimes, it will take several events

in order to piece together a pattern that reveals the answer.